Obtaining Structural Parameters from STEM−EDX Maps of Core/ShellNanocrystals for OptoelectronicsJacob T. Held,† Katharine I. Hunter,‡ Nabeel Dahod,§ Benjamin Greenberg,‡

Danielle Reifsnyder Hickey,† William A. Tisdale,§ Uwe Kortshagen,‡ and K. Andre Mkhoyan*,†

†Department of Chemical Engineering and Materials Science and ‡Department of Mechanical Engineering, University of Minnesota,Minneapolis, Minnesota 55455, United States§Department of Chemical Engineering, Massachusetts Institute of Technology, Cambridge, Massachusetts 02139, United States

*S Supporting Information

ABSTRACT: Characterization efforts of core/shell and core/multishell nanocrystals have struggled to quantitativelyevaluate the interface width between the core and shellmaterials despite its importance in their optoelectronicproperties. Here, we demonstrate a scanning transmissionelectron microscopy (STEM) method for measuring the radialelemental composition of two spherical core/shell nanocrystalsystems, Ge/Si core/shell and CdSe/CdS/ZnS core/double-shell nanocrystals. By fitting model-based radial distributionsof elements to measured STEM−energy-dispersive X-ray(EDX) maps, this method yields reliable and accuratemeasurements of interface broadening as well as core and shell sizes, surface roughness, and the fraction of core material inthe shell. The direct evaluation of the structural parameters is an important step toward improving the synthesis of core/shellnanocrystals and optimizing their optoelectronic properties.

KEYWORDS: core, shell, nanocrystal, interface, STEM, EDX, characterization

■ INTRODUCTION

Semiconductor nanocrystals (NCs) exhibit many interestingand useful size-dependent optoelectronic properties1,2 and havegained increasing applications in fields ranging from opticaldisplays3,4 to biological imaging.5−7 They are often coveredwith shells to protect them from oxidation and surface trapstates8−10 and to better control the optoelectronic11−14 andcatalytic15,16 properties through induced strain and bandalignment between the core and shell. The properties ofthese core/shell NCs can be further tuned by controlling thecore and shell shape, size, and composition.17−19 In addition tothese parameters, the sharpness of the interface between thecore and shell also affects the properties of the NCs.20−25

Therefore, it is important to reliably measure and modify thisinterface. However, measuring the width of the core/shellinterface remains a major challenge in the characterization ofthese NCs.Raman spectroscopy has been used to detect interfacial

mixing in core/shell nanostructures,26−28 but it is not applicablefor all core/shell NCs and only yields ensemble-average valuesfrom relatively large quantities of NCs. Scanning transmissionelectron microscopy (STEM) combined with energy-dispersiveX-ray (EDX) spectroscopy or electron energy-loss spectrosco-py, on the other hand, presents the unique opportunity toanalyze individual NCs at the atomic scale with a wide varietyof elemental compositions.14,29−31 It should be noted that

tomography in transmission electron microscopy (TEM) canalso achieve atomically accurate reconstructions of someNCs.32−34 However, this approach is inapplicable when thedifference between the atomic numbers (ΔZ) of the elementsin the core and shell is small, which is common in core/shellNCs.Earlier reports demonstrated that, despite limitations due to

electron beam damage, reliable STEM−EDX maps and lineprofiles can be obtained from individual core/shell NCs,showing their core/shell structure.31,35 Some attempts to obtainelemental distributions from nanomaterials using STEM−EDXhave also been made.36−38 However, these studies do notprovide a method for the systematic and quantitativedetermination of structural information from core/shell NCsthat is applicable to a wide variety of elemental compositionsand is capable of accurately evaluating the core/shell interfacewidth of individual NCs.In this study, we show that, given a well-defined (in this case,

spherical) NC geometry, it is possible to quantify with highcertainty the radial elemental composition of a core/shell NCfrom a single STEM−EDX map and, most importantly, toestimate the sharpness of the interface between the core and

Received: December 27, 2017Accepted: January 11, 2018Published: January 11, 2018

Article

www.acsanm.orgCite This: ACS Appl. Nano Mater. 2018, 1, 989−996

© 2018 American Chemical Society 989 DOI: 10.1021/acsanm.7b00398ACS Appl. Nano Mater. 2018, 1, 989−996

shell. We demonstrate a quantitative analysis of the elementaldistribution by fitting a model to STEM−EDX maps of theNCs. The analysis is first performed on plasma-grown Ge/Sicore/shell NCs.14 Then, we demonstrate the practicalapplication of this analysis to widely studied CdSe/CdS/ZnScore/double-shell NCs.8,24,39

■ METHODSSynthesis of Ge/Si Core/Shell NCs. The details of the process of

producing Ge/Si core/shell NCs from a capacitively couplednonthermal plasma were described previously.14 In brief, this core/shell NC growth was achieved in a single flow-through plasma reactorby first injecting the core precursor and, after its depletion, injectingthe shell precursor further downstream. An argon plasma wasgenerated by the application of 50 W forward power at 13.56 MHzto a pair of copper electrodes encircling a 1-in.-o.d. glass tube. Theprimary gas feed, consisting of argon (20 sccm), dihydrogen (25sccm), and GeCl4 vapor (50 mTorr partial pressure), entered thereactor upstream of the plasma. Within the plasma, germanium (Ge)NCs nucleate and grow until the GeCl4 precursor is depleted. Furtherdownstream, the secondary gas feed, consisting of SiH4 (0.2 sccm)diluted in argon (9.8 sccm), was injected into the plasma through astainless steel toroidal inlet with 16 evenly spaced 1 mm holes aroundthe circumference of the reactor tube. Within the plasma, SiH4decomposes and grows as an epitaxial silicon (Si) shell onto the Gecores. The total pressure in the reactor chamber was held constant at 2Torr by an orifice placed downstream of the plasma. NCs weredeposited directly onto thin carbon with holey support TEM grids byacceleration through this orifice and impaction onto the grid.Synthesis of CdSe/CdS/ZnS Core/Shell/Shell NCs. CdSe cores

were prepared via modification of a previously reported method.40 Thesynthesis was conducted in a nitrogen environment through the use ofan air-free Schlenk line. First, cadmium oxide (80 mg) was combinedwith stabilizing organic ligands, tri-n-octylphosphine oxide (TOPO; 6g) and octadecylphosphonic acid (562 mg), and degassed at 120 °C.The resulting solution was then heated to 300 C under nitrogen andleft until it became clear (approximately 2 h). The temperature wasramped to 320 °C, and tri-n-octylphosphine (TOP; 1.5 mL) wasinjected to replenish the noncoordinated ligand in solution. Thetemperature was then ramped to the growth temperature (355 °C),and a solution of selenium (Se) coordinated with TOP was swiftlyinjected to create the nucleation burst that initiates NC growth. NCswere purified via precipitation, first by the addition of acetone toremove excess TOPO and then by the addition of a 2:1 methanol/butanol solution to precipitate the NCs. The final product wasredispersed in an organic solvent and stored in a nitrogen glovebox.The epitaxial growth of CdS and ZnS capping shells on previously

prepared CdSe NCs proceeded via modification of the protocolpublished by Boldt et al.24 Stock solutions of zinc oleate and cadmiumoleate were prepared as outlined in that protocol. Briefly, in a nitrogenenvironment (via Schlenk line), a dispersion of previously preparedCdSe NCs (100 nmol) was dissolved in a coordinating organic solvent[3 mL of oleylamine and 3 mL of octadecene (ODE)], and thetemperature was raised to 310 °C. When the temperature reached 280°C, cadmiums (cadmium oleate) and sulfur (octanethiol) precursorsdissolved in ODE were added dropwise via a syringe pump. Thequantity of the cadmium precursor was used to control the shellthickness and is calculated using the core size, desired thickness, andlattice constant of wurtzite CdS. A 1.5-fold excess of thiol was used,and the injection rate was varied such that one monolayer of CdS wasgrown per hour. After the injection was complete, the solution wasmaintained at 200 °C, while organic ligands were replenished via thedropwise injection of 1 mL of oleic acid. Lingering volatiles in thesolution were pulled off via vacuum (30 min), and the temperature wasthen raised to 280 °C. When the temperature reached 240 °C,dropwise injection of the zinc (zinc oleate) and sulfur (octanethiol)precursors in ODE was started. The quantity of the precursor requiredand the rate of injection were determined as above for the CdS shellpreparation. After the precursors were added, the reaction mixture was

allowed to cool to room temperature. The CdSe/CdS/ZnS core/double-shell NCs were purified via precipitation by the addition ofacetone until the solution became turbid. The precipitate wasredispersed and purified at least three times, or until the resultingsolution remained clear. Purified dispersions were stored in a nitrogenglovebox.

An aliquot of the CdSe/CdS/ZnS NC dispersion was diluted 10times and drop-cast onto ultrathin carbon with holey support TEMgrids. To protect the particles from beam damage during STEM−EDXmap acquisition, the grids were then coated with Al2O3 by atomic layerdeposition (ALD) using a Cambridge Nanotech/Ultratech SavannahS2000 reactor and via a procedure similar to that described byThimsen et al.41 First, the NCs were heated under a dinitrogen flow at180 °C and 0.3 Torr for 30 min. Then, ∼4 nm Al2O3 coatings weredeposited at 180 °C using 40 cycles of ALD. Sequentially, each cycleconsisted of a 0.1 s trimethylaluminum (TMA) pulse, a 30 s dinitrogenpurge, a 0.1 water pulse, and a 30 s dinitrogen purge. The maximumpressure was 3 Torr during TMA pulses and 400 Torr during waterpulses, and the minimum pressure was 0.3 Torr during dinitrogenpurges.

STEM−EDX Measurements. All NCs were analyzed using anaberration-corrected FEI-Titan G2 60-300 sanning tensmissionelectron microscope equipped with a Super-X energy-dispersive X-ray spectrometer located at the University of Minnesota.42,43 STEM−EDX experiments were performed at 60 kV STEM beam voltage witha 25 mrad convergence angle, a 125 pA beam current, and a 68 mradhigh-angle annular-dark-field (HAADF) detector inner angle. EDXspectral images were acquired with frame-by-frame drift correctionthrough the Bruker ESPRIT 1.9 software with a dwell time of 3 μs/pixel and a pixel resolution of 0.04 nm/pixel for the Si/Ge sample and0.06 nm/pixel for the CdSe/CdS/ZnS sample. For the Si/Ge sample,the Si and Ge K-edges were background-subtracted and integrated,producing the spectral images. In the CdSe/CdS/ZnS NC sample, theS K-edge and Se, Cd, and Zn L-edges were used.

■ RESULTS AND DISCUSSION

The Ge/Si NCs used in this study were predominantlyspherical in shape. Therefore, a spherical, concentric, core/shellNC model was used as the basis for the simulations. Takingadvantage of this symmetry, the distributions of elementswithin the NCs were expressed as one-dimensional (1D)functions of the spherical radius that were integrated along thez axis to produce 1D projected distributions of elements.Likewise, the experimental two-dimensional (2D) STEM−EDXmaps were converted into 1D intensity profiles by radiallyaveraging the EDX counts for each element around the centerof the particle, as illustrated in Figure 1, allowing a directcomparison between the experimental data and simulateddistributions.Spherical radial concentration distributions were constructed

using error functions to approximate core/shell interfacebroadening. This functional form was chosen because thesolution to a spherically symmetrical interdiffusion modelbetween a core and shell converges to these equations whenlittle diffusion has occurred. This approximation also has theadvantages of yielding intuitive results in the form of a standarddeviation of each distribution and of being computationallyinexpensive relative to the diffusion model. The similarity of thediffusion model and error function approximation for the core/shell systems studied here is demonstrated in the SupportingInformation (SI). By using only the error function model, nodistinction is made between the “roughness” and “interdiffu-sion” of the core and shell species. At the scale of individualatoms, as discussed here, this distinction is physically unclear.The error function model was also used as an approximationfor the surface roughness.

ACS Applied Nano Materials Article

DOI: 10.1021/acsanm.7b00398ACS Appl. Nano Mater. 2018, 1, 989−996

990

When this approximation of the elemental distribution isapplied to a simple core/shell system and the radius of the coreis denoted as rcore and the width of the interface between thecore and shell as σi (the standard deviation of the representativeerror function), the radial distribution of the core material A isexpressed as

σσ

= = −−⎡

⎣⎢⎢⎛⎝⎜

⎞⎠⎟⎤⎦⎥⎥c r c r r

r r( ) ( , , )

12

1 erf2A core core i

core

i (1)

Likewise, the radial distribution of the shell material B isexpressed as

σ σ

σσ

=

= −−

−⎡⎣⎢⎢

⎛⎝⎜

⎞⎠⎟⎤⎦⎥⎥

c r c r r r

r rc r r

( ) ( , , , , )

12

1 erf2

[1 ( , , )]

B shell core i shell s

shell

score core i

(2)

where the outer radius of the shell is denoted as rshell and theroughness of the outer surface of the shell as σs. This system oftwo coupled equations was then used to determine all fourparameters describing the elemental distribution of the core/shell NC: the radii of the core and shell (rcore and rshell) and thewidths of the core/shell interface and the surface roughness (σiand σs). Before eqs 1 and 2 were applied as fitting functions tothe experimental EDX data set to deduce these parameters,they were integrated along the STEM beam axis, z, to reflectthe fact that experimental EDX maps are 2D projections of 3DNCs. Therefore, the two fitting functions are

∫=−∞

+∞f r c r z( ) ( ) dA 2D A (3)

∫=−∞

+∞f r c r z( ) ( ) dB 2D B (4)

The pair of functions (3) and (4) can, in principle, describeany STEM−EDX map obtained from spherical core/shell NCswith a nonsharp interface and surface, provided that littleinterdiffusion has occurred. However, in the systems studiedhere, NC synthesis introduced a nonnegligible concentration ofthe core material in the shells. To account for this, eqs 1 and 2were modified to

σ σ σ= +c r c r r pc r r r( ) ( , , ) ( , , , , )Ge core core i shell core i shell s (5)

σ σ= −c r p c r r r( ) (1 ) ( , , , , )Si shell core i shell s (6)

where p is the fraction of the residual core material, in this caseGe, in the Si shell.

Figure 1. Schematic representation of the acquisition of an STEM−EDX map of a core/shell NC and radially averaged data from such amap. The (top) ideal case represents the radially averaged data of aperfectly spherical core and shell with a sharp interface and surface,mapped with a δ-functional probe. The (bottom) realistic caseincludes a diffuse interface and surface and a finite probe size.

Figure 2. Ge/Si core/shell NC analysis. (a) HAADF−STEM image and EDX maps of a Ge/Si core/shell NC. The HAADF image of the NC takenafter EDX map acquisition shows Z contrast between the Ge core and Si shell. The composite EDX map of Ge (green) and Si (red) and individualEDX maps of Ge and Si demonstrate a clear differentiation between the core and shell material. (b) Radially averaged EDX data with correspondingfit, with core and shell signals scaled separately. (c) Parameters corresponding to the optimized fit.

ACS Applied Nano Materials Article

DOI: 10.1021/acsanm.7b00398ACS Appl. Nano Mater. 2018, 1, 989−996

991

The effects of the finite size of the STEM probe and itsbroadening should also be taken into consideration. In STEM,the probe that scans the sample has a finite size, defined by theconvergence angle and voltage as well as the effective sourcesize.44,45 Scattering of the probe by atoms in the sample resultsin additional broadening as it propagates through the sample.Multislice-based simulations46 of the experimental setup showthat, with beam broadening included, the effective size of theSTEM probe in this study was approximately 0.18 nm. Detailsof these simulations are presented in the SI. Therefore, thefitting functions (3) and (4) were convoluted with a Gaussianfunction with a full width at half-maximum of 0.18 nm toincorporate the effective size of the STEM probe before fittingto the EDX maps.The fitting error was minimized using the Nelder−Mead

method47 to obtain all parameters describing the core and shellstructure of the core/shell NCs: rcore, σi, rshell, σs, and p. Thefitting error was calculated as

∑= −=N

r I IEr1

[ ( )]j

N

j j1

2Dfit exp 2

j(7)

where N is the total number of data points in the intensityprofiles and rj, Ij

fit, and Ijexp are the radius, fitted intensities, and

experimental intensities, respectively. Here, the residual wasmultiplied by the 2D radius to account for the number of pixelsencountered at each radius increasing proportionally to 2πr2Dj

.This also accounts for the higher noise at low radii for radiallyaveraged EDX data sets, which can be seen in Figures 2 and 4.After optimization of all parameters in each data set, the fittinguncertainty of each parameter was defined as the maximumchange in the parameter that was necessary to increase thefitting error by 5%. The complete flowchart of this analysis canbe found in the SI.To prepare the experimental data for analysis, an ellipse was

first fit to the STEM−EDX maps of each NC to determine thecenter and major/minor axis lengths and to ensure that thecenters of the maps of each element were coincident (details ofthis analysis are provided in the SI). The average position of thefit centers was then set as the overall center of the NC, andeach EDX map was radially averaged around this point. Tomaintain the spherical approximation of the NC geometry, NCswith an aspect ratio of 1.05 or greater were discarded, alongwith those with the centers for each element that varied bymore than 0.4 nm, or 10 pixels. Data sets that met these criteriawere further analyzed.A uniform background of the Si EDX signal due to residual Si

on the grids from NC synthesis was subtracted. The Sibackground was found by averaging the signal outside the NCfor each EDX map. An example of this is shown in the SI. Nosignificant Ge background was detected. Simulated projecteddistributions of the elements in spherical NCs were then fit tothe radially averaged and background-subtracted experimentaldata sets. One example of the full analysis is presented in Figure2.The fitting procedure for the Ge/Si system resulted in a very

good agreement between the experimental data and model, asshown in Figure 2. The radial distribution of the elementsdetermined through this analysis is shown in Figure 3. Thebroadening at the core/shell interface as well as the surfaceroughness was determined to be 0.5 nm; this is comparable tothe lattice parameter of the NC (0.56 nm), which was shown tobe slightly strained by Hunter et al.14 Because these values arise

from a combination of interdiffusion at the interface, roughness,asymmetry, and nonconcentricity of the core/shell NCs, theyrepresent the total contribution of these effects to thedivergence from a perfect spherical core/shell system. In thiscase, the Ge core had an aspect ratio of 1.01 and the Si shellhad an aspect ratio of 1.02, contributing 0.02 and 0.08 nm tothe interface and surface roughness, respectively. Additionally,there was 0.2 nm offset between the core and shell centers,contributing 0.1 nm to the core/shell interface broadeningmeasurement. No significant faceting was observed in theSTEM images of the NCs studied here. The remaininginterface and surface broadening widths (approximately 0.4 nmfor the core/shell interface and 0.32 nm for the NC surface)arise from roughness/interdiffusion. Analyses of several otherNCs are presented in the SI.Additionally, analysis showed a uniform 2.5% Ge content in

the Si shell. If this concentration of Ge in the shell were due tointerdiffusion, it would be reflected in σGe/Si, and there wouldbe a significantly greater concentration of Si in the Ge core dueto the much larger volume of the shell (approximately 10 timesthe core volume). It should also be noted that Ge K-edge X-rays (9.87 keV) are of higher energy than Si K-edge X-rays(1.74 keV), such that this Ge signal could not be produced viasecondary fluorescence. While the opposite could happen(producing Si EDX counts), this interaction is extremelyimprobable because of the small size of the NCs (∼10 nm)relative to the attenuation length (∼120 μm).48 The addition ofanother parameter to account for the shell material in the corefit was less than 0.5% and did not result in a significantly betterfit. Therefore, we conclude that there was residual Ge precursorin the plasma at the point of Si precursor injection during theNC synthesis.14

Aside from some surface alterations, little change in theradially averaged HAADF profiles obtained before and afterEDX acquisition was observed, demonstrating that minimalbeam damage occurred during the experiment (for details, seethe SI). While most core/shell NC systems exhibit little

Figure 3. (a) Radial (spherical) distribution of Ge and Si within theNC shown in Figure 2, obtained from fitting. (b) HAADF imageintensity profile fitted with a “Z-contrast” approximation using eq 8.

ACS Applied Nano Materials Article

DOI: 10.1021/acsanm.7b00398ACS Appl. Nano Mater. 2018, 1, 989−996

992

contrast between the core and shell materials, the largedifference in the atomic number, and therefore contrastbetween Ge and Si, offered a unique opportunity to test theresults from EDX analysis against HAADF−STEM images ofthe same particles. Therefore, for further confirmation of themodel, the obtained projected distributions of the elementswere fit to the HAADF signal according to

= +α αI r A f r Z f r Z( ) [ ( ) ( ) ]ADF2D Ge 2D Ge Si 2D Si (8)

where f Ge(r2D) and f Si(r2D) are the projected radial distributionsof Ge and Si (eqs 3 and 4), respectively, determined by fittingto EDX maps. ZGe and ZSi are the atomic numbers of Ge and Si,and A and α are the fitting parameters. The parameter Aprovides scaling for the experimental ADF intensity, and itsvalue is not meaningful for this analysis. The best fitcorresponded to α = 1.4, which lies within the range ofquantitative STEM models for Z contrast in HAADF−STEMimaging (1.2−1.8)49−51 and matches well with the moreaccurate prediction of 1.5 ± 0.1 by Wang et al.51 for theexperimental parameters used here. The result of this analysis isshown in Figure 3. It should be noted that, in principle, fullanalysis could be performed using only the HAADF−STEMdata, such as that shown in Figure 3, but the fitting procedurewould still require all five of the fitting parameters tabulated inFigure 2, plus α from eq 8. The reliability of such a six-parameter fit to a single data set (as opposed to five parameterssplit between two EDX map data sets) would be questionable.Additionally, the nontrivial background contrast from thecarbon support, carbon contamination, and presence of ligandsor ALD coating would complicate HAADF analysis evenfurther. HAADF−STEM analysis also requires prior knowledgeof the elemental composition, which is implicitly determined byEDX mapping.In addition to the added complexity of a second shell, the

CdSe/CdS/ZnS core/double-shell NCs studied here were

significantly smaller and more beam-sensitive than the Ge/SiNCs. Because of residual ligands from the synthesis procedure,carbon contamination via oligomerization of loose ligandsunder the STEM beam also posed a challenge to STEM−EDXanalysis. When the crystals already deposited on the TEM gridswere coated with alumina via ALD to a thickness of 4 nm (40deposition cycles), most of the experimental challenges weremitigated; the ligands were driven off during ALD, and thealumina coating protected the crystals during analysis. It shouldbe noted that thinner ALD coatings of only 3−4 depositioncycles also reduced carbon contamination while leaving a moreeven background for imaging. However, these thin coatings didlittle to protect the NCs, which damaged rapidly under theSTEM beam.During EDX map acquisition, the alumina film suffered some

beam damage, contributing a significant textured background tothe HAADF images. While the major sources of contrast in theimage remained stationary, allowing reliable drift correctionduring acquisition, the damaged alumina film made itimpractical to quantitatively compare the before/afterHAADF images of the NCs in the same way as the Ge/Sisystem. However, the underlying NC sustained minimaldamage during the EDX map acquisition based on theconsistent shape, dimensions, and lattice features of theunderlying NC observed in the before/after HAADF images(see the SI).The same radial-averaging preparation procedure as that

described for the Ge/Si NCs was applied to the EDX maps ofthe CdSe/CdS/ZnS NCs. Here, the sulfur (S) and cadmium(Cd) maps were used to determine the center of the NCbecause they provided the highest total counts of the elementsin this system, and therefore the best signal-to-noise ratio.The core/double-shell CdSe/CdS/ZnS NC was modeled as

two coupled core/shell distributions: one representing theanionic Se/S core/shell and the other representing the cationicCd/Zn core/shell, with the surface radius and roughness shared

Figure 4. Application of the fitting procedure to a CdSe/CdS/ZnS core/double-shell NC. (a) HAADF−STEM image and EDX maps of a CdSe/CdS/ZnS core/double-shell NC. A three-pixel Gaussian blur has been applied to all EDX maps for visual clarity. (b and c) Radially averaged EDXdata for the anion (Se/S) and cation (Cd/Zn) distributions and corresponding fits, scaled separately. (d) Parameters corresponding to the optimizedfit. (e) Radial (spherical) distribution of each element within the NC, obtained from the fitting.

ACS Applied Nano Materials Article

DOI: 10.1021/acsanm.7b00398ACS Appl. Nano Mater. 2018, 1, 989−996

993

between the two distributions. A schematic diagram of thistreatment is provided in the SI. Thus, this model required fourcoupled elemental distribution equations, one representingeach element, and three more parameters than the Ge/Si core/shell NC: the dimensions and broadening of the additionalinterface and the concentration of the core material in thesecond shell. One example of this analysis is presented inFigure 4.The results of this analysis indicated a relatively sharp (0.3

nm) CdSe/CdS interface and a slightly broader (0.5 nm) CdS/ZnS interface. The outer surface roughness was 0.5 nm.Because the thickness of the ZnS shell was only 0.6 nm, thebroadening of the CdS/ZnS interface and the outer surfaceroughness overlap over the majority of the ZnS shell, as isshown in Figure 4e. In this sample, there was a largeconcentration of the core material in the outer shells, withthe S shell composed of about 20% Se and the Zn shellcontaining about 40% Cd. We believe that this is due tosolvation and redeposition of the material during shell growthas well as the presence of unreacted cadmium oleate during theaddition of the ZnS shell.Because these NCs were smaller and yielded fewer total EDX

counts from each element before nonnegligible damage wasobserved, the uncertainties in each fit parameter werepercentagewise larger than those in the Ge/Si sample analysisbut still represent only a fraction of the crystal unit cell (forexample, aCdS = 0.416 nm), which, from a structuralcharacterization point of view, is quite good. The significantnoise and poorer fits observed in the Se and Zn EDX databelow 1−2 nm radii arise from the low number of EDX countsfor these elements. However, EDX data from the comple-mentary elements (S and Cd), which are present in higherconcentrations in the NCs, are less noisy. Because each of thefit curves represents the optimized fit of the pair of elements,this provides extra fitting stability even when some elementshave relatively noisy EDX data sets. Despite the complexity ofthe double-shell system and the practical limitations of dataacquisition, the method, nonetheless, yields quantitative resultsfor the size and broadening of both core/shell interfaces.

■ CONCLUSIONSIn this study, we demonstrated a model-based fitting methodfor the quantitative compositional analysis of STEM−EDXmaps of spherical core/shell NCs. Simulated projected radialdistributions of the elements were fit to the corresponding EDXintensity profiles, yielding values for the core and shell sizes,interface broadening, surface roughness, and fraction of corematerial in the shell. The method was demonstrated for twospherical NC systems: simpler Ge/Si core/shell and morecomplex CdSe/CdS/ZnS core/double-shell NCs. The methodallows an evaluation of the interface widths with precision ashigh as 0.1−0.2 nm.Although this analysis was limited to spherical core/shell

NCs, it could be expanded to more complex NC geometries byanalyzing the full 2D maps rather than radially averaging thedata into one dimension. Such a procedure would increase thenumber of simultaneous fitting parameters. Therefore, a bettersignal-to-noise ratio in the EDX maps would be required tomaintain the same certainty in the fit.This method is readily applicable to a wide variety of

elemental compositions of core/shell NCs, provided that eachelement exhibits resolvable EDX spectral features. Additionally,it relies on the relative EDX intensities rather than the absolute

intensities and therefore does not require detector calibrationto EDX standards. Finally, it provides important localinformation specific to an individual NC but could be readilyautomated to quantify the properties of numerous core/shellNCs for improved statistical analysis of a sample of NCs. Giventhe influence of the interface broadening, composition, andsurface structure on the optoelectronic properties of core/shellNCs and the current lack of tools to evaluate these parameters,this characterization approach may serve as an important stepin optimizing the synthesis of core/shell systems for use inoptoelectronic devices.

■ ASSOCIATED CONTENT*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acsanm.7b00398.

Evaluation of a diffusion-based model, details on probebroadening, flowchart for the analysis procedure, detailson center determination and background subtraction,supplemental data sets and analyses, evaluation ofelectron beam damage, and schematic description ofthe core/double-shell analysis approach (PDF)

■ AUTHOR INFORMATIONCorresponding Author*E-mail: mkhoyan@umn.edu. Tel: 612-625-2059. Fax: 612-626-7246.

ORCIDJacob T. Held: 0000-0003-3864-4314Katharine I. Hunter: 0000-0001-8639-5012Benjamin Greenberg: 0000-0002-5013-5996William A. Tisdale: 0000-0002-6615-5342K. Andre Mkhoyan: 0000-0003-3568-5452NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSThis project was supported by the MRSEC program of theNational Science Foundation under Award DMR-1420013.STEM analysis was performed in the College of Science andEngineering Characterization Facility of the University ofMinnesota, which receives partial support from the NSFthrough the MRSEC program. N.D. and W.A.T. weresupported by the Division of Chemistry and Division ofMaterials Research, U.S. National Science Foundation, underAward 1452857. D.R.H. was supported by C-SPIN, one of sixcenters of STARnet, a Semiconductor Research Corporationprogram, sponsored by MARCO and DARPA.

■ REFERENCES(1) Norris, D. J.; Bawendi, M. G. Measurement and Assignment ofthe Size-Dependent Optical Spectrum in CdSe Quantum Dots. Phys.Rev. B: Condens. Matter Mater. Phys. 1996, 53 (24), 16338−16346.(2) Mangolini, L. Synthesis, Properties, and Applications of SiliconNanocrystals. J. Vac. Sci. Technol., B: Nanotechnol. Microelectron.:Mater., Process., Meas., Phenom. 2013, 31 (2), 20801.(3) Bourzac, K. Quantum Dots Go on Display. Nature 2013, 493(7432), 283−283.(4) Shirasaki, Y.; Supran, G. J.; Bawendi, M. G.; Bulovic,́ V.Emergence of Colloidal Quantum-Dot Light-Emitting Technologies.Nat. Photonics 2013, 7 (1), 13−23.

ACS Applied Nano Materials Article

DOI: 10.1021/acsanm.7b00398ACS Appl. Nano Mater. 2018, 1, 989−996

994

(5) Bruchez, M., Jr.; Moronne, M.; Gin, P.; Weiss, S.; Alivisatos, A. P.Semiconductor Nanocrystals as Fluorescent Biological Labels. Science1998, 281 (5385), 2013−2016.(6) Medintz, I. L.; Uyeda, H. T.; Goldman, E. R.; Mattoussi, H.Quantum Dot Bioconjugates for Imaging, Labelling and Sensing. Nat.Mater. 2005, 4 (6), 435−446.(7) Michalet, X.; Pinaud, F. F.; Bentolila, L. A.; Tsay, J. M.; Doose, S.;Li, J. J.; Sundaresan, G.; Wu, A. M.; Gambhir, S. S.; Weiss, S. QuantumDots for Live Cells, in Vivo Imaging, and Diagnostics. Science(Washington, DC, U. S.) 2005, 307 (5709), 538−544.(8) Pust, P.; Schmidt, P. J.; Schnick, W. A Revolution in Lighting.Nat. Mater. 2015, 14 (5), 454−458.(9) Wang, F.; Liu, X. Recent Advances in the Chemistry ofLanthanide-Doped Upconversion Nanocrystals. Chem. Soc. Rev. 2009,38 (4), 976−989.(10) Schar̈tl, W. Current Directions in Core-Shell NanoparticleDesign. Nanoscale 2010, 2 (6), 829−843.(11) Balet, L. P.; Ivanov, S. A.; Piryatinski, A.; Achermann, M.;Klimov, V. I. Inverted Core/Shell Nanocrystals Continuously Tunablebetween Type-I and Type-II Localization Regimes. Nano Lett. 2004, 4(8), 1485−1488.(12) Reiss, P.; Protier̀e, M.; Li, L. Core/Shell SemiconductorNanocrystals. Small 2009, 5 (2), 154−168.(13) Smith, A. M.; Mohs, A. M.; Nie, S. Tuning the Optical andElectronic Properties of Colloidal Nanocrystals by Lattice Strain. Nat.Nanotechnol. 2009, 4 (1), 56−63.(14) Hunter, K. I.; Held, J. T.; Mkhoyan, K. A.; Kortshagen, U. R.Nonthermal Plasma Synthesis of Core/Shell Quantum Dots: StrainedGe/Si Nanocrystals. ACS Appl. Mater. Interfaces 2017, 9 (9), 8263−8270.(15) Strasser, P.; Koh, S.; Anniyev, T.; Greeley, J.; More, K.; Yu, C.;Liu, Z.; Kaya, S.; Nordlund, D.; Ogasawara, H.; Toney, M. F.; Nilsson,A. Lattice-Strain Control of the Activity in Dealloyed Core-Shell FuelCell Catalysts. Nat. Chem. 2010, 2 (6), 454−460.(16) Sasaki, K.; Wang, J. X.; Naohara, H.; Marinkovic, N.; More, K.;Inada, H.; Adzic, R. R. Recent Advances in Platinum MonolayerElectrocatalysts for Oxygen Reduction Reaction: Scale-up Synthesis,Structure and Activity of Pt Shells on Pd Cores. Electrochim. Acta2010, 55 (8), 2645−2652.(17) Chen, W.-T.; Yang, T.-T.; Hsu, Y.-J. Au-CdS Core−ShellNanocrystals with Controllable Shell Thickness and PhotoinducedCharge Separation Property. Chem. Mater. 2008, 20 (23), 7204−7206.(18) Ghosh Chaudhuri, R.; Paria, S. Core/Shell Nanoparticles:Classes, Properties, Synthesis Mechanisms, Characterization, andApplications. Chem. Rev. 2012, 112 (4), 2373−2433.(19) Zhang, F.; Che, R.; Li, X.; Yao, C.; Yang, J.; Shen, D.; Hu, P.; Li,W.; Zhao, D. Direct Imaging the Upconversion Nanocrystal Core/Shell Structure at the Subnanometer Level: Shell Thickness Depend-ence in Upconverting Optical Properties. Nano Lett. 2012, 12 (6),2852−2858.(20) McBride, J.; Treadway, J.; Feldman, L. C.; Pennycook, S. J.;Rosenthal, S. J. Structural Basis for Near Unity Quantum Yield Core/Shell Nanostructures. Nano Lett. 2006, 6 (7), 1496−1501.(21) Cragg, G. E.; Efros, A. L. Suppression of Auger Processes inConfined Structures. Nano Lett. 2010, 10 (1), 313−317.(22) García-Santamaría, F.; Brovelli, S.; Viswanatha, R.;Hollingsworth, J. A.; Htoon, H.; Crooker, S. A.; Klimov, V. I.Breakdown of Volume Scaling in Auger Recombination in CdSe/CdSHeteronanocrystals: The Role of the Core-Shell Interface. Nano Lett.2011, 11 (2), 687−693.(23) Bae, W. K.; Padilha, L. A.; Park, Y.-S.; McDaniel, H.; Robel, I.;Pietryga, J. M.; Klimov, V. I. Controlled Alloying of the Core-ShellInterface in CdSe/CdS Quantum Dots for Suppression of AugerRecombination. ACS Nano 2013, 7 (4), 3411−3419.(24) Boldt, K.; Kirkwood, N.; Beane, G. A.; Mulvaney, P. Synthesis ofHighly Luminescent and Photo-Stable, Graded Shell CdSe/Cd X Zn1− X S Nanoparticles by In Situ Alloying. Chem. Mater. 2013, 25 (23),4731−4738.

(25) Maiti, S.; Debnath, T.; Maity, P.; Ghosh, H. N. Tuning theCharge Carrier Dynamics via Interfacial Alloying in Core/Shell CdTe/ZnSe NCs. J. Phys. Chem. C 2016, 120 (3), 1918−1925.(26) Tschirner, N.; Lange, H.; Schliwa, A.; Biermann, A.; Thomsen,C.; Lambert, K.; Gomes, R.; Hens, Z. Interfacial Alloying in CdSe/CdSHeteronanocrystals: A Raman Spectroscopy Analysis. Chem. Mater.2012, 24 (2), 311−318.(27) Dzhagan, V. M.; Valakh, M. Y.; Milekhin, A. G.; Yeryukov, N.A.; Zahn, D. R. T.; Cassette, E.; Pons, T.; Dubertret, B. Raman- andIR-Active Phonons in CdSe/CdS Core/Shell Nanocrystals in thePresence of Interface Alloying and Strain. J. Phys. Chem. C 2013, 117(35), 18225−18233.(28) Todescato, F.; Minotto, A.; Signorini, R.; Jasieniak, J. J.; Bozio,R. Investigation into the Heterostructure Interface of CdSe-BasedCore−Shell Quantum Dots Using Surface-Enhanced Raman Spec-troscopy. ACS Nano 2013, 7 (8), 6649−6657.(29) McBride, J. R.; Kippeny, T. C.; Pennycook, S. J.; Rosenthal, S. J.Aberration-Corrected Z-Contrast Scanning Transmission ElectronMicroscopy of CdSe Nanocrystals. Nano Lett. 2004, 4 (7), 1279−1283.(30) Rosenthal, S. J.; McBride, J.; Pennycook, S. J.; Feldman, L. C.Synthesis, Surface Studies, Composition and Structural Character-ization of CdSe, Core/Shell, and Biologically Active Nanocrystals. Surf.Sci. Rep. 2007, 62 (4), 111−157.(31) Abel, K. A.; Boyer, J.-C.; Andrei, C. M.; van Veggel, F. C. J. M.Analysis of the Shell Thickness Distribution on NaYF 4 /NaGdF 4Core/Shell Nanocrystals by EELS and EDS. J. Phys. Chem. Lett. 2011,2 (3), 185−189.(32) Midgley, P. A.; Weyland, M. 3D Electron Microscopy in thePhysical Sciences: The Development of Z-Contrast and EFTEMTomography. Ultramicroscopy 2003, 96 (3−4), 413−431.(33) Bals, S.; Casavola, M.; van Huis, M. A.; Van Aert, S.; Batenburg,K. J.; Van Tendeloo, G.; Vanmaekelbergh, D. Three-DimensionalAtomic Imaging of Colloidal Core-Shell Nanocrystals. Nano Lett.2011, 11 (8), 3420−3424.(34) Van Aert, S.; Batenburg, K. J.; Rossell, M. D.; Erni, R.; VanTendeloo, G. Three-Dimensional Atomic Imaging of CrystallineNanoparticles. Nature 2011, 470 (7334), 374−377.(35) Kim, S. Y.; Jeong, J. S.; Mkhoyan, K. A.; Jang, H. S. DirectObservation of the Core/double-Shell Architecture of Intense Dual-Mode Luminescent Tetragonal Bipyramidal Nanophosphors. Nano-scale 2016, 8 (19), 10049−10058.(36) Lauhon, L. J.; Gudiksen, M. S.; Wang, D.; Lieber, C. M. EpitaxialCore−shell and Core−multishell Nanowire Heterostructures. Nature2002, 420 (6911), 57−61.(37) Gan, L.; Heggen, M.; Rudi, S.; Strasser, P. Core-ShellCompositional Fine Structures of Dealloyed Pt xNi 1-X Nanoparticlesand Their Impact on Oxygen Reduction Catalysis. Nano Lett. 2012, 12(10), 5423−5430.(38) Hadar, I.; Philbin, J. P.; Panfil, Y. E.; Neyshtadt, S.; Lieberman,I.; Eshet, H.; Lazar, S.; Rabani, E.; Banin, U. Semiconductor SeededNanorods with Graded Composition Exhibiting High Quantum-Yield,High Polarization, and Minimal Blinking. Nano Lett. 2017, 17 (4),2524−2531.(39) Yalcin, A. O.; Goris, B.; van Dijk-Moes, R. J. a; Fan, Z.; Erdamar,A. K.; Tichelaar, F. D.; Vlugt, T. J. H.; Van Tendeloo, G.; Bals, S.;Vanmaekelbergh, D.; Zandbergen, H. W.; van Huis, M. a. Heat-Induced Transformation of CdSe-CdS-ZnS Core-Multishell QuantumDots by Zn Diffusion into Inner Layers. Chem. Commun. (Cambridge,U. K.) 2015, 51 (16), 3320−3323.(40) Peng, Z. A.; Peng, X. Formation of High-Quality CdTe, CdSe,and CdS Nanocrystals Using CdO as Precursor. J. Am. Chem. Soc.2001, 123 (1), 183−184.(41) Thimsen, E.; Johnson, M.; Zhang, X.; Wagner, A. J.; Mkhoyan,K. A.; Kortshagen, U. R.; Aydil, E. S. High Electron Mobility in ThinFilms Formed via Supersonic Impact Deposition of NanocrystalsSynthesized in Nonthermal Plasmas. Nat. Commun. 2014, 5, 5822.(42) Odlyzko, M. L.; Held, J. T.; Mkhoyan, K. A. Atomic BondingEffects in Annular Dark Field Scanning Transmission Electron

ACS Applied Nano Materials Article

DOI: 10.1021/acsanm.7b00398ACS Appl. Nano Mater. 2018, 1, 989−996

995

Microscopy. II. Experiments. J. Vac. Sci. Technol., A 2016, 34 (4),41603.(43) Jeong, J. S.; Odlyzko, M. L.; Xu, P.; Jalan, B.; Mkhoyan, K. A.Probing Core-Electron Orbitals by Scanning Transmission ElectronMicroscopy and Measuring the Delocalization of Core-LevelExcitations. Phys. Rev. B: Condens. Matter Mater. Phys. 2016, 93(16), 165140.(44) LeBeau, J. M.; Findlay, S. D.; Allen, L. J.; Stemmer, S.Quantitative Atomic Resolution Scanning Transmission ElectronMicroscopy. Phys. Rev. Lett. 2008, 100 (20), 206101.(45) LeBeau, J. M.; Findlay, S. D.; Wang, X.; Jacobson, A. J.; Allen, L.J.; Stemmer, S. High-Angle Scattering of Fast Electrons from CrystalsContaining Heavy Elements: Simulation and Experiment. Phys. Rev. B:Condens. Matter Mater. Phys. 2009, 79 (21), 214110.(46) Kirkland, E. J. Advanced Computing in Electron Microscopy, 2nded.; Springer Science & Business Media, 2010.(47) Nelder, J. A.; Mead, R. A Simplex Method for FunctionMinimization. Comput. J. 1965, 7 (4), 308−313.(48) Hubbel, J. H.; Seltzer, S. M. Tables of X-Ray Mass AttenuationCoefficients and Mass Energy-Absorption Coefficients, version 1.4;National Institute of Standards and Technology: Gaithersburg, MD,2004. Available at https://www.nist.gov/pml/x-Ray-Mass-Attenuation-Coefficients (accessed Nov 3, 2017).(49) Kirkland, E. J.; Loane, R. F.; Silcox, J. Simulation of AnnularDark Field Stem Images Using a Modified Multislice Method.Ultramicroscopy 1987, 23 (1), 77−96.(50) Pennycook, S. J. Z-Contrast Stem for Materials Science.Ultramicroscopy 1989, 30 (1−2), 58−69.(51) Wang, Z. W.; Li, Z. Y.; Park, S. J.; Abdela, A.; Tang, D.; Palmer,R. E. Quantitative Z-Contrast Imaging in the Scanning TransmissionElectron Microscope with Size-Selected Clusters. Phys. Rev. B:Condens. Matter Mater. Phys. 2011, 84 (7), 73408.

ACS Applied Nano Materials Article

DOI: 10.1021/acsanm.7b00398ACS Appl. Nano Mater. 2018, 1, 989−996

996